At 16:10 21/06/97 -0400, Donald H. Mitchell wrote:
>Francis Leboutte wrote:
>
>> ..I have
>> been satisfied with the AllegroStore 1.1.x versions: very easy to use
>> (it
>> was very easy to port an existing application to Allegrostore), a bit
>> slow
>> and not very well integrated in the development environment (e.g. no
>> online
>> help for the AS functions). The final application has proved to be
>> very
>> reliable.
>>
>> However the last version (1.2) seems to be a true disaster even if the
>>
>> performance has been improved ..
>
>As one of the original VARs to sponsor the astore development, I have a
>very different experience. I've found 1.2 to be vastly superior to 1.1
>in speed, function, and reliability. We could not deliver our
>application under 1.1 due to reliability and performance problems. We
>are delivering it under 1.2. There are differences, but they are well
>known ...
 
Apparently we have a rather different experience with 1.2. Actually I have
switched back to AS 1.1.12. Maybe you are using Allegrostore on a Unix
machine (I use AS with ACL4W 3.0.2 and WIndows NT4), maybe most of my
problems are du to the inverse functions I use.
Anyway, in my opinion, the addition of AS (1.1 or 1.2) to ACL4W has the
effect to make the development *tool* clearly less robust.

>> Current version is 1.2 . Too many crashes (application exception ->
>> access
>> violation), undefined bugs ("hard error") and other various bugs (e.g.
>> in
>> the inverse functions)...
>
>I removed all uses of inverse fns in our application a long time ago
>because the performance was so poor. I was thinking about adding them
>now that 1.2 advertises great improvements. Perhaps I should avoid them.
>
>As far as hard errors and access violations, have you checked the
>aclxdump.txt files to see if they captured a backtrace?  We have
>experienced some access violations (about one every 100 hours of use)
>but have not been able to detect a pattern nor get reliable backtraces
>in the aclxdump files.  Stack overflows are mostly likely problems in
>your code: the aclxdump files should help.
>

I don't think so. I experiment these stack overflows with AS 1.2 and the
code that has been used several months in production (or this code plus
very small additions).
Frankly I don't want to explore the aclxdump.txt file (actually I have no
time for that). Of course I have sent this file to the support when requested.

Francis


--
Francis Leboutte, Algorithme, Rue de la Charrette 141, 4130 Tilff, Belgium
<skynet.be at f.leboutte>     <acm.org at leboutte>               
http://users.skynet.be/algo    t&fax: +32-(0)4-388.35.28